In the early 1980s a new class of thiazolidinedione-based insulin sensitizers (TZDs) was reported. These compounds, termed glitazones, were later shown to be potent and selective PPARγ agonists and are now widely used for the treatment of type 2 diabetes. PPARs modulate multiple aspects of lipid and carbohydrate metabolism and are therefore relevant targets for treating several important aspects of type 2 diabetes and the metabolic syndrome. However, some TZDs were associated with cases of liver toxicity, which resulted in the withdrawal of troglitazone from the market. This and other side effects such as fluid retention and weight gain, associated with the commercially available rosiglitazone and pioglitazone, demonstrate the need to develop new compounds with improved efficacy and reduced side-effect profiles.
In this report we presented a new class of PPARα/γ modulators, the 2-BABAs. A set of representative compounds, termed BVT.13, BVT.762, and BVT.763, were shown by x-ray crystallography to utilize a novel binding epitope not involving the classical agonist-characteristic interactions. Exemplified compounds within the 2-BABA family displayed agonistic activity in a PPARγ cell-based reporter gene assay and failed to antagonize the rosiglitazone-induced activity at an approximate EC50 concentration of 30 nM. In vivo, the specific PPARγ agonist BVT.13, displayed antidiabetic effects in ob/ob mice by significantly reducing the plasma levels of glucose, insulin, triglycerides, and free fatty acids.
Our current understanding of the ligand-induced transcription of nuclear receptors originates from mutational studies, highlighting the importance of the C-terminal part of the ligand binding domain, which has been denoted AF2 (32). The increasing number of LBD structures has more recently contributed to the understanding of this mechanism on a molecular level. Taken together, these observations resulted in the conclusion that NR agonists stabilize the active conformation of the receptor primarily by locking helix 12 (AF2) in a conformation that allows coactivators to bind to a specific recognition site, mainly formed by helices 3, 4, and 12. Direct interactions between the agonist and amino acid residues of helix 12 have been suggested as the structural basis for transcriptional activation of many receptors (22, 24). One exception is the retinoid X receptor where the natural ligand 9-cis-retinoic acid only interacts indirectly with helix 12 via residues Cys-269 and Ala-272 of helix 3 and Trp-305 of helix 5 (33), localized in the vicinity of helix 12. The previously reported holoPPAR complexes share a conserved network of hydrogen bonds between the ligand and the side chains of residues His-323, His-449, and Tyr-473 (PPARγ nomenclature), which therefore has been suggested as crucial for the ligand-dependent activation of the PPARs (5, 21, 23, 34). Surprisingly, hydrogen bonding to this triad is not conserved for the 2-BABA compounds, which suggests that these interactions are not necessary for the induction of transcriptional activation by PPARα and -γ. Furthermore, the 2-BABAs are situated at the entrance of the active site and do not interact directly with helix 12. There are three unrelated crystallographic apo structures of PPARγ (21, 26), of which two closely resemble the agonistic structures in terms of helix 12 position and interactions, thus giving no hints of conformational changes related to ligand activation. In the third structure helix 12 adopts a different conformation, slightly protruding away from the LBD, and could therefore more easily be interpreted as an inactive state. Furthermore, NMR studies of the PPAR and the PPARγ LBDs have shown that the apo forms of the receptors are in an equilibrium of conformations rather than adopting one single stable conformation (23, 35, 36). Altogether, this suggests that the unliganded LBD of PPARs can adopt the active conformation as well as a number of inactive conformations. Despite the lack of direct interactions with AF2, the binding of the 2-BABA compounds seem to push the equilibrium of conformational states toward an active state, which suggests an alternative activation mechanism for PPARγ by an overall stabilization of the LBD. This activation mechanism is further supported by the observation that a partial PPARγ agonist, nTZDpa, only partially stabilizes the receptor (35). As judged by superpositioning of the holoPPAR complexes, all agonists induce highly similar conformations of the LBD despite their different interactions with the receptor (Fig. 5). Large variations are shown in three clustered regions, H2 to the N-terminal part of H3, the N-terminal part of H7, and the C-terminal part of H11 to the N-terminal part of H12. These regions show higher B factors than the LBD in general, and the differences seen could be because of a higher degree of mobility in these regions. It has been reported that PPAR agonists differ in their ability to induce recruitment of specific coactivators (37). Again the PPARγ complexes do not give a straight answer to this selectivity, and it is possible that selectivity is generated by subtle differences in the dynamics of the LBD not detectable in these crystal structures.
The cell-based reporter gene assay revealed that all three substances, BVT.13, BVT.762, and BVT.763, are in fact potent activators of PPARγ. BVT.762 and BVT.763 were also shown to activate PPARα, whereas BVT.13 failed to bind and transcriptionally activate PPARα (25
). Structural alignments of the PPAR subtypes revealed one noteworthy amino acid difference in the vicinity of the heterocyclic substituent where residue Gly-284 in PPARγ has been substituted by a cysteine in PPARα and an arginine in PPARδ. The cysteine and the arginine side chains may restrict the space available for short and bulky substituents, providing an explanation for the selectivity of BVT.13.
In accordance with the in vitro data of PPARγ selective agonist activity, the administration of BVT.13 led to a significant reduction in plasma levels of glucose, insulin, triglycerides, free fatty acids, and cholesterol (Fig. 4 and data not shown). Furthermore, a significant increase in body weight was observed for both the BVT.13- and the rosiglitazone-treated animals compared with the vehicle-treated, although no difference in food intake was observed (data not shown). Weight increase is commonly associated with PPARγ agonist treatment and generally associated with the pharmacological activity mediated through PPARγ agonism (38). The ob/ob mouse, as a model for type 2 diabetes, is known to respond well to the PPARγ agonistic TZDs, which was also confirmed here by the rosiglitazone-treated group. The kinetic study using 100 mg/kg of BVT.13, yielded an approximate average steady-state unbound concentration of 1 µM. This dose corresponds to an approximate EC50 concentration, as determined for BVT.13 in the cell-based reporter gene assay. This suggests that at least the two higher dose groups were sufficiently exposed to BVT.13 to respond through PPARγ activation. We believe that the antidiabetic effects observed are likely because of PPARγ activation by BVT.13.
We concluded that the 2-BABA binding mode can be used to design isoform-specific PPAR modulators with biological activity in vivo and that direct interaction with helix 12 is not a necessity for transcriptional activation. With respect to the large ligand binding pockets of the PPARs (1250-1500 Å3), our observations present new possibilities for the design of PPAR modulators interacting with the receptor through binding sites distal to the classical activation site. This might also have implications for the transcriptional activation of other orphan NRs.